1. Introduction
Conventional chest radiography has limited sensitivity in detecting lung abnormalities at the early stage of pneumonia [
1,
2]. Manifestations associated with neutropenia are subtle and difficult to determine with plain chest X-rays [
3,
4]. As a result, CT of the chest is the imaging method of choice for the prompt detection of radiological signs consistent with pulmonary infections and is frequently performed in patients with hematologic malignancies. Nonetheless, the radiation dose levels of chest CT are substantial and raise radiation concerns for these high-risk patients who are vulnerable to fatal infections triggered by therapy for neutropenia [
5]. Moreover, repeated scanning to follow up on pneumonia progression or to monitor patient response to treatment also increases the radiation dose [
6,
7,
8]. It is, therefore, essential that radiation doses are in compliance with the As Low As Reasonably Achievable (ALARA) principle to ensure acceptable diagnostic quality and reasonable image noise levels. The most common dose reduction strategy is tube current reduction and is preferable in Low-Dose CT (LDCCT) protocols [
9,
10]. Consequently, LDCCT most often yields higher image noise levels and thus to a loss of low-contrast spatial resolution impairing the overall image quality [
11]. Even though LDCCT scanning is capable of radiation dose reduction by approximately 1/4 of the standard dose, the radiation-induced cancer risk for doses less than 100 mSv is unpredictable [
12,
13,
14,
15]. Since the introduction of iterative reconstruction (IR) algorithms, quantum or statistical image noise can be removed systemically [
16]. Today, LDCCT is capable of rendering images of sufficient image quality for the detection of lung abnormalities. In addition, when IR algorithms are implemented, they have the potential to either further reduce image noise or radiation dose depending on the diagnostic requirement [
17,
18].
To our knowledge, there are no previous studies comparing data sets from low-dose and standard-dose CT examinations of the chest with the use of specific pulmonary infection criteria and statistical-based IR algorithms for adult patients with hematologic malignancies. The aim of our study was to investigate the image quality and the diagnostic performance of LDCCT for the diagnosis and monitoring of pulmonary infections in patients with hematologic malignancies.
4. Discussion
Pulmonary infections, especially invasive mold lung infections, including invasive pulmonary aspergillosis and pulmonary mucormycosis, are a significant cause of morbidity and mortality among neutropenic patients with hematological malignancies, and/or hematopoietic stem cell transplantation [
27]. Early diagnosis and timely initiation of an appropriate antifungal therapy are of paramount importance, as diagnostic delays are associated with increased mortality [
28]. Neutropenia blunts the immune response and the inflammatory process, making diagnostic modalities, such as chest X-ray, ineffective due to low sensitivity [
7]. The modern approach to the neutropenic patient with fever is the diagnostic-driven or pre-emptive approach, when the decision to start antifungal therapy is not based solely on the presence of fever not responding to antibiotics but on diagnostic modalities including serial screening of serum galactomannan, aspergillus PCR and serial high-resolution CT scans, on demand [
29]. Pioneering work from von Eiff M. et al. and Caillot D. et al. has shown that regular chest CT scanning in febrile neutropenic patients with invasive pulmonary aspergillosis (IPA) can significantly reduce the overall mortality rate by 50% [
7,
30,
31].
The high diagnostic value of high-resolution CT scans in patients with hematological malignancies resulted in an increased frequency of CT examinations, which has tripled over the past 15 years, contributing to almost 60% of the collective radiation dose from medical exposures [
32]. Therefore, concerns have been raised about the frequent irradiation of these vulnerable patients, as it can lead to long-term complications whose effects remain unknown [
11,
26]. In response, international radiation protection authorities have launched dose reduction campaigns, particularly focusing on CT and interventional imaging [
33,
34].
In recent years, vendors have focused on developing CT systems that incorporate noise reduction algorithms to optimize patient radiation dose while maintaining diagnostic value. These algorithms have been designed to minimize the impact on diagnostic accuracy, ensuring that the quality of the images remains high despite the reduced radiation dose. This development has been ongoing since 2011, with continuous improvements aimed at achieving a balance between dose reduction and preserving diagnostic value in CT imaging [
35]. These algorithms systematically remove image noise and are commonly used in ≤1 mm CT imaging to mitigate artifacts (streak, beam hardening, and photon starvation) caused by the dense shoulder girdle [
36]. However, the implementation of IR algorithms in lung CT examinations, especially for immunocompromised patients, is a subject of debate. Indistinct findings such as interstitial disease, small nodules, and ground glass opacities (GGO) are best visualized with higher spatial resolution and edge enhancement [
37].
It is widely recognized that high-resolution kernels used for lung CT examinations increase image noise [
38,
39,
40]. When combined with thinner slices (<1 mm) and low mAs settings of LDCCT (40 mAs), the reduced radiation dose can further increase image noise and degrade overall image quality. IR algorithms have been shown to reduce image noise, but they exhibit low-contrast detectability similar to Filtered Back Projection (FBP) with higher radiation exposure reductions (>30%) with sufficient noise reduction impartial to image pixelization (blocky appearance) [
19,
41].
In this work, we reduced the radiation dose of CT scans, and we assessed the diagnostic performance of low-dose CT scans in neutropenic patients with hematological malignancies. We modified the most common dose reduction parameter, the effective tube current, while keeping the reconstruction and image presentation parameters consistent with standard-dose chest CT (SDCCT). Both protocols employed IR algorithms (SAFIRE
TM) unlike previous studies that focused on comparisons between low-dose chest CT (LDCCT) and chest radiography or between different CT techniques [
3,
18,
19,
42,
43,
44]. Alternatively, our study integrated SAFIRE
TM S3 with the strength set at level—S3 (strength: S1–S5) to produce images.
We found that LDCCT achieved a dose reduction of 47% with satisfactory image quality and acceptable image noise compared to SDCCT (
Figure 4). However, the diagnostic performance of LDCCT was lower, underestimating significant radiologic findings associated with pulmonary infections in neutropenic patients, especially consolidation and GGO. LDCCT detected consolidation and GGO in less than 1/3 of cases compared to SDCCT. The detection rate for consolidation and ground glass opacity was influenced by sex, age, and BMI. On the other hand, the diagnostic performance was similar between the two protocols for other radiologic findings such as cavitation in nodules, diffuse interlobular septal thickening, pleural effusion, pericardial effusion, and lymphadenopathy.
A previous study by Hae et al. concluded that ultra-low-dose CT (ULDCT) (Deff: 0.60mSv ± 0.15) with FBP provides acceptable image quality and 63.6% sensitivity for diagnosing pulmonary infections in febrile neutropenic and hematologic malignancy patients. However, the final diagnosis was verified through additional clinical information, laboratory findings, and follow-up chest X-rays [
19]. Another study reported that unenhanced LDCCT with IR generated images of improved quality and reduced image noise compared to FBP. They suggested that lesion conspicuity is greatly improved with increasing ASIR strength [
43].
Kubo et al. demonstrated that low-dose chest CT (LDCCT) using 50 mAs (effective dose: 3.57 mSv) is effective in detecting various pulmonary abnormalities, including emphysema, ground glass opacities (GGO), reticular opacity, micronodules, bronchiectasis, honeycomb, and nodules larger than 5 mm. In comparison, standard-dose chest CT (SDCCT) using 150 mAs (effective dose: 10.7 mSv) showed similar diagnostic capabilities but with a threefold higher radiation dose. Despite the higher dose in their LDCCT, our study enabled a higher GGO sensitivity at 64% compared to their 49%. Kubo et al. suggested the need for further research on the accurate classification of interstitial pneumonia in patients with a high prevalence of interstitial lung disease [
45].
Another study supported the potential of up to 65% dose reduction using SAFIRE
TM-based chest CT image reconstruction. This approach resulted in images with reduced image noise by 31%–59% and provided good diagnostic confidence compared to conventional chest CT (SDCCT) with filtered back projection (FBP). Although the investigators highlighted the superiority of SAFIRE
TM algorithms over FBP, they concluded that these algorithms may not be directly applicable for lung nodule follow-up, lung cancer screening, and bronchiectasis evaluation, where low-dose FBP protocols offer improved visibility of small anatomical structures [
19].
Although CT has been the cornerstone for early diagnosis of pulmonary fungal infections among patients with hematological malignancies, biomarkers play a pivotal role as well. A recent prospective study has shown that a preemptive antifungal strategy including twice weekly serum galactomannan screening and CT scan on demand is safe and effective. In addition, this strategy is not associated with an increased risk of invasive fungal infection, and reduces greatly the use of antifungals [
29]. Similarly, Picardi M et al., have shown in a retrospective study that among high-risk patients receiving prophylaxis with posaconazole the application of serial serum beta D-Glucan tests and an aggressive strategy of early chest CT scans allows early diagnosis of breakthrough pulmonary aspergillosis, even before the appearance of halo sign and serum galactomannan increase [
46].
The limitations of our study include a lack of follow-up on patient progress and the effect of LDCCT on prognosis and survival rate, compared to SDCCT. Additionally, both data sets were obtained using only a single IR algorithm strength setting, specifically SAFIRE
TM S3 as incorporated in the departmental protocol for standard-dose chest CT (SDCCT). However, Kalra et al. proposed that using a higher strength setting, such as SAFIRE
TM S4, could provide images of acceptable diagnostic quality despite the pixelized appearance [
20]. This limitation might have impacted the ability to fully assess the diagnostic performance of low-dose chest CT (LDCCT) in detecting significant pulmonary infection findings, such as consolidation and ground glass opacification in the lungs of patients with hematologic malignancies. Another limitation is that the effect of antifungal therapy on the findings of follow-up LDCCT compared to initial SDCCT was not taken into account. The differences in diagnostic ability might be partly due to the effect of a successful antifungal therapy. However, radiological findings of pulmonary fungal infections, especially nodules, do not abate so quickly after the initiation of antifungal therapy, as they might persist for several weeks.
Further research is advocated, whereby a minimal increase in mAs compared to LDCCT while maintaining a submSv effective radiation dose or a modification in SAFIRETM strength, may facilitate the detection of pulmonary infection-specific radiologic findings in neutropenic patients with underlying hematologic conditions.